Automated and computationally efficient joint motion analysis using low quality fluoroscopy images

The kinematic analysis of joint motion has proven to significantly improve orthopedic surgeries by enhancing surgery assessment, prosthesis design, and pathology diagnosis. This analysis is usually performed by a three to two dimensional (3D/2D) registration of the 3D bone model to a 2D radiographic video from C-arm fluoroscopy imaging machines. However, the practicality of such analysis is undermined due to lengthy and user-dependent 3D/2D image registration algorithms and the high cost of C-arm fluoroscopy imaging machines. Mini C-arm fluoroscopic machines are a more affordable alternative, but the low quality of their images has hindered their use in this application. In this thesis, we propose a novel 3D/2D image registration algorithm for the kinematic analysis of joint bones using mini C-arm fluoroscopy machines with significantly improved speed, despite their low quality images. This method performs a fast registration through a novel algorithm for quick and high-quality generation of digitally reconstructed radiographs (DRR), which is the bottleneck in such processes. Moreover, the dependency of the results on the user has been reduced as a new feature-based registration algorithm replaced the previously manual initialization phase of the process. This algorithm is able to reach the true registration from within 90 degrees of it, which is a substantial improvement over the existing methods. In addition, our algorithm performs the registration in significantly reduced time due to a smaller number of generated DRRs.Ph.D.Includes bibliographical referencesby Soheil Ghafuria